On Tue, 27 Feb 2024 at 03:00, Doug Anderson <diand...@chromium.org> wrote:
>
> Hi,
>
> On Mon, Feb 26, 2024 at 4:37 PM Dmitry Baryshkov
> <dmitry.barysh...@linaro.org> wrote:
> >
> > On Sat, 24 Feb 2024 at 00:40, Hsin-Yi Wang <hsi...@chromium.org> wrote:
> > >
> > > This series is a follow up for 1a5e81de180e ("Revert "drm/panel-edp: Add
> > > auo_b116xa3_mode""). It's found that 2 different AUO panels use the same
> > > product id. One of them requires an overridden mode, while the other
> > > should
> > > use the mode directly from edid.
> > >
> > > Since product id match is no longer sufficient, EDP_PANEL_ENTRY2 is
> > > extended
> > > to check the crc hash of the entire edid base block.
> >
> > Do you have these EDIDs posted somewhere? Can we use something less
> > cryptic than hash for matching the panel, e.g. strings from Monitor
> > Descriptors?
>
> We could try it if need be. I guess I'm worried that if panel vendors
> ended up re-using the panel ID for two different panels that they
> might also re-use the name field too. Hashing the majority of the
> descriptor's base block makes us more likely not to mix two panels up.
> In general it feels like the goal is that if there is any doubt that
> we shouldn't override the mode and including more fields in the hash
> works towards that goal.
My main concern is that hash (or crc) is a non-obvious thing: even if
we have EDID in the commit message, it takes some effort to calculate
the value. If for any reason we decide to change the hashed bytes
(e.g. by dropping any of the fields) it will be an error-prone process
to update existing hash values. On the contrary, the 'strings' are
easy to check / compare without any additional tools.
>
> I guess one thing that might help would be to make it a policy that
> any time a panel is added to this list that a full EDID is included in
> the commit message. That would mean that if we ever needed to change
> things we could. What do you think?
Definitely, that sounds like a good idea.
> That being said, if everyone thinks that the "name" field is enough,
> we could do it. I think that in the one case that we ran into it would
> have been enough...
Yes, I'd suggest using the string unless at some point we see two
panels sharing both panel_id and names.
--
With best wishes
Dmitry
